The impact of inaccurate property data in process modelling

Abstract

It is widely accepted that high-quality thermophysical property data are essential for the accurate modelling of chemical processes, both in the conceptual design phase of process development and in the optimization of existing processes. Unfortunately, over the last decades, the chemical industry has experienced the closure of many physical property data laboratories. This trend is not limited to experimental facilities, also applied thermodynamic groups have been significantly downsized or sometimes extinguished. Process engineers and modelling experts often take over the tasks of evaluating and testing the thermodynamic property packages used in their models. From the simulation industry, we see that it is increasingly common that users request out-of-the-box property data that can be used directly in their simulations. Estimation methods are sometimes being used without a proper acknowledgment of their limitations and associated uncertainties. We believe it is, therefore, important to raise awareness of how large the impact of potential property data errors on process modelling can be, more specifically on the modelling of typical downstream unit operations. In this work, we provide practical insights on this issue, by revisiting textbook examples and by delving into real-life industrial cases we have encountered over the years. Critical considerations on the direct use of data from large databanks and estimation methods are also presented. The last is particularly relevant, with an increasing number of research groups working on the development of machine learning methods as means to generate massive amounts of property data.